WO2019202252A1 - Reliable detection of an object by a vehicle - Google Patents

Abstract

The invention relates to a method for processing data relating to the assisted driving of a current vehicle (Vk), for the detection of an object (On) likely to modify the driving of said vehicle (Vk), said method comprising the steps of: acquisition (S1), by means of at least two sensors (C1, C2, C3), of data relating to said object; calculation (S2) of a probability of the existence of said object, at least from the data of the sensors (C1, C2, C3); detection (S5) of said object according to at least the probability of existence and the number of sensors alone having acquired data indicating the presence of the object; and generation (S6) of a driving assistance instruction according to said detection.

Description

 Reliable detection of an object by a vehicle

The present invention claims the priority of the French application 1853507 filed April 20, 2018 whose content (text, drawings and claims) is here incorporated by reference. The present invention belongs to the field of assistance in driving a vehicle. It relates in particular to a method of reliably detecting an object likely to modify the driving of the vehicle.

It is particularly advantageous in the case of an autonomous motor vehicle in a bottling situation. "Vehicle" means any type of vehicle such as a motor vehicle, a moped, a motorcycle, a rail vehicle, etc. By assistance to the driving of a vehicle is meant any automated process capable of assisting the driving of the vehicle. The method may thus consist of partially or totally steering the vehicle or providing any type of assistance to the natural person driving the vehicle. The assistance devices implementing these methods can, for example, facilitate parking or the rear steps, or detect obstacles, especially in front of their vehicle, or illuminate an obstacle detected in front of their vehicle (sometimes called "marking light" function) , or correct the trajectory of their vehicle according to the marking delimiting the traffic lanes used, or regulate the speed of their vehicle according to a set point provided by their driver or a speed limit in force on the route of borrowed traffic.

In order to be able to function, most of these assistance devices need to be fed at least with information representative of the environment of their vehicle. This information is usually provided by, or derived from, data from exteroceptive sensors, such as ultrasonic sensors, cameras, radars or lidars.

The higher the level of autonomy of the pipe, the higher the level of security of the information transmitted by the sensors and processed by the components in charge of calculating the instructions. driving must be high. In particular, the detection of objects in the vehicle environment and likely to require a change of direction, braking or acceleration of the vehicle must be extremely reliable.

To do this, it has been proposed to multiply the number of sensors to ensure redundancy in the information acquired by the sensors. Thus, a false detection (false positive) or a lack of false detection (false negative) by a sensor must be able to be detected by another sensor in charge of detecting the same type of object.

This solution, however, has several disadvantages. In the first place, despite the multiplication of detection means, errors persist. When sensor readings are obviously faulty, the detection is based on one or a small number of sensors, which can still lead to erroneous detections. In addition, the multiplication of sensors has a significant cost and complicates the layout and management of the vehicle components.

The present invention improves the situation.

For this purpose, a first aspect of the invention relates to a data processing method relating to the assisted driving of a current vehicle, for the detection of an object likely to modify the driving of the vehicle, comprising the steps of:

 • acquisition by at least two data sensors relating to the object;

 • calculation of a probability of existence of the object, at least from the data of the sensors;

 • detection of the object in function, at least:

 o the probability of existence;

 o the number of sensors that have acquired, on their own, data indicating the presence of the object;

Generating a driving assistance instruction according to said detection. The testing of consolidated information, the probability of existence, with raw sensor data, the number of sensors having detected the object alone, significantly increases the reliability of the detection of the object.

In particular, the number of sensors is an information which, statistically, is very discriminating and effective for detecting the errors of calculation of the probabilities of existence. Conversely, in cases where the detection decision is primarily made according to the number of sensors, the probability of existence brings an additional fineness in the decision making detection.

The combination of these two criteria therefore has the effect of making reliable detection decision making, without having to increase the redundancy of the sensors. It is therefore the best compromise between reliability of the detection and complexity / costs controlled.

The term "number of sensors having acquired, by themselves, data indicating the presence of the object" the number of sensors having detected an object autonomously. This means, for example, that if only one in three sensors has detected an object, the data acquired by that sensor indicates, without having to be mapped to other data from other sensors (for example), than object is present.

In one embodiment, the method comprises, prior to the step of calculating the probability of existence, a step of:

• verification that said object is in the path of said vehicle; in the case where the object is not on the path, generating a non-detection instruction.

The computation resources mobilized for the detection of the object are thus advantageously reduced. The term "non-detection instruction" means any type of data from which it can be deduced that no object has been detected.

In particular, in one embodiment, the object is considered in the path of said vehicle if the object is moving towards the path. The term "moving towards the trajectory" means any movement of the object likely to bring it into an area where he expected the vehicle to go. In one embodiment, the detection of the object is also a function of the reaction to said object of a vehicle preceding the current vehicle.

Additional relevant information and not requiring additional sensor is thus advantageously used to further enhance the reliability of the detection of the object. Indeed, objects, including static objects, detected by sensors of the current vehicle and which did not cause reaction of the previous vehicle have almost certainly been detected by mistake by the current vehicle.

"Previous vehicle" means any vehicle preceding the current vehicle. For example, if the running vehicle has a uniform straight forward motion, the previous vehicle is the one in front of the current vehicle.

"Vehicle reaction" is understood to mean any type of physical or digital reaction of the preceding vehicle. For example, braking, acceleration, changing direction, sending an emergency message, changing the driving style, etc. are vehicle reactions. In particular, in one embodiment, the reaction to said object of the preceding vehicle is determined by implementing substeps of: i. acquisition by the sensors of the current vehicle of a history of reactions of the preceding vehicle;

 ii. determining from the history of the reaction of the preceding vehicle to said object.

In a particular embodiment, the reaction of the vehicle preceding a braking or acceleration action.

In one embodiment, the detected object is a static object.

In another embodiment, the method comprises, after the detection step, steps of:

• intelligence of a statistical base of false detection of object, including the substeps of: generating a false detection input when an object, called a noise object, has been detected by said detection step but it has also been detected that a vehicle preceding said current vehicle has not reacted in response to this object;

 o storing the characteristics of said noise object;

 • statistical determination of false object detection, comprising, when an object has been detected in the detection step, the sub-steps of:

 o determination of the characteristics of the object;

 o search for characteristics common to the detected object and noise objects present in the statistical database;

 o in case of correspondence, generation of a false detection instruction.

In another embodiment, the method further comprises a step of determining a lateral trajectory of the current vehicle. In this embodiment, the object is not detected if said sensors reveal that the object is not in the path of the current vehicle.

A second aspect of the invention is directed to a computer program comprising instructions for implementing the method according to the first aspect of the invention, when these instructions are executed by a processor.

A third aspect of the invention relates to a device for processing data relating to the assisted driving of a current vehicle, for the detection of an object likely to modify the driving of said vehicle, comprising:

At least two data sensors relating to said object;

 A processor arranged to perform the operations of:

 calculating a probability of existence of said object, at least from the data of the sensors;

 o detection of said object in function, at least:

^■ the probability of existence;

^■ the number of sensors that have acquired, on their own, data indicating the presence of the object; o generating a driving assistance instruction according to said detection.

A fourth aspect of the invention is a vehicle comprising the data processing device according to the third aspect of the invention.

Other characteristics and advantages of the invention will appear on examining the detailed description below, and the attached drawings in which: FIG. 1 illustrates an application context of the invention; FIG. 2 illustrates a data processing method according to the invention; - Figure 3 illustrates a microcontroller, according to one embodiment of the invention.

The invention is described below in its non-limiting application to the case of an autonomous motor vehicle in a bottling situation. Other applications such as a bus on a dedicated lane or a motorcycle on a country road are also possible.

Figure 1 illustrates a motor vehicle V _k . The motor vehicle V _k is preceded by the preceding vehicle V _k -i, both present on an RA track, for example a highway lane. The method according to the invention relates to the reliable detection of an object O _n . Sensors C1, C2 and C3 are present on the vehicle V _k to acquire data capable of translating the presence of the object O _n . The sensors C1, C2 and C3 are, for example, exteroceptive sensors, for example ultrasonic sensors, cameras, radars, a GPS (for "global positionning system", a global positioning system in French), a sensor of movement or lidars. These sensors are connected to a data processing device D relating to the assisted driving described hereinafter with reference to FIG.

Figure 2 is a diagram showing the steps of the method according to one embodiment of the invention.

In step S1, the sensors C1, C2 and C3 acquire data relating to the object O _n . Such data are able to translate the presence of the object O _n . For example, for a radar sensor, data reflecting the detection of an object would correspond to a sensor instruction indicating the presence of an object in the sensor field and possibly an approximate distance between the object and the sensor.

An object can only be detected by a sensor, two sensors or by the three sensors. In this case, the sensor data is acquired but does not reflect the presence of an object.

The sensor data can be consolidated for example by identifying the type of object detected by the sensor or sensors. For example, an image analysis can lead to classification of objects (static / mobile), attributes (positions, speeds, etc.), typing (car, motorcycle, etc.), etc.

When different objects Oi, O2,. . .. One is detected, object-by-object processing can be performed. In this situation, the objects can be processed successively in order of detection, according to their class, attribute or typing, etc. Simultaneous treatment, at least partially, of the objects can also be implemented.

In one embodiment, an additional step (not shown) for determining the trajectory of the vehicle V _k is implemented. From the data, typically consolidated, sensors, data relating to the vehicle or data relating to the driver of the vehicle (age, years of license, etc.), a lateral trajectory (direction to the left or right) is calculated .

The data relating to the vehicle are for example: speed, wheel orientation, characteristics (speed, temperature, etc.) of the engine, energy reserves (gasoline, electricity, etc.), positioning (typically by means of a GPS, for Global Positioning System in French global positioning system, integrated), a state of tire wear, etc.

Among the objects detected by the sensors, it is possible that some are out of the lateral path of the vehicle. In one embodiment, this means that any object likely to be in the path of the vehicle, regardless of the longitudinal path (towards the front or rear) of the vehicle, is taken into account.

In another embodiment, assumptions are made to define the longitudinal trajectory of the vehicle. These assumptions are for example a speed interval equal to [speed V _k - 30 km / h; speed V _k + 20 km / h] over a fixed distance, or an evolutionary speed depending on the distance. In this other embodiment, only the objects on the trajectory determined according to said hypotheses are taken into account.

In another embodiment, an area in which objects are taken into account is defined. For example, depending on the guide mode of the vehicle, the selection of objects will be on a conical zone whose initial width is the track, which decreases with distance while following the curvature of the track (guidance mode called "tracking of track "). For another guidance mode, called "vehicle flow monitoring", the zone corresponds to the estimated lane width around the target trajectory.

Adjacent lane objects can also be taken into account in two particular cases:

When an insertion path is detected in the vicinity: identified from cartographic information, ground markings and the presence of objects on adjacent lanes;

When the trajectory prediction made it possible to evaluate that the trajectory of the vehicle of the adjacent lane collided with the trajectory of the current vehicle (so-called "aggressive cut-in" situation).

The method then comprises, prior to the step S2 of calculating a probability of existence, a step of:

• verification that said object is in the path of said vehicle; in the case where the object is not on the path, generating a non-detection instruction.

In particular, in one embodiment, the object is considered in the path of said vehicle if the object is moving towards the path. These are typically objects on adjacent channels. In step S2, a probability P _n of existence of the object O _n is calculated at least from the data acquired by the sensors. The probability of existence is for example calculated by weighting data received from the sensors. The weighting is for example a function of the location of the object, the type of sensor concerned, the class, the attributes or the typing of the object. The data relating to the vehicle or its trajectory can also be taken into account. Other data, for example received from outside via a radio frequency internet connection, can also be taken into account.

A detection step S5 is then implemented.

In one embodiment, the detection step is made as a function of the probability P _n of existence of the object O _n and the number of sensors having acquired, by themselves, data indicating the presence of the object O _n .

Different situations are possible to take the decision to withhold or not the detection of the object O _n .

In one embodiment, existence probability intervals are defined. At couples [number of sensors detected; probability interval] are associated with driving assistance instructions for the vehicle.

For example, for a static object uncertainly detected by two out of three sensors:

• if the probability of existence is greater than 80%: braking instruction on the object;

• if the probability of existence is between 50 and 80%: limited braking for a predetermined duration (for example 4 seconds) then braking according to the evolution of the number of sensors and the probability after the predetermined duration;

• if the probability of existence is less than 50%: no action taken. In another embodiment, the reaction of the vehicle V _k -i is also taken into account during the step of detecting an object O _n by the vehicle V _k .

To do this, a history of the reactions of the previous vehicle is stored in step S3. The reactions of the preceding vehicle can be collected by the sensors C1, C2, C3 and / or transmitted from the vehicle V _k -i to the vehicle V _k .

When the previous vehicle is tracked from the data acquired by the sensors C1, C2, C3, a treatment, for example when the previous vehicle is detected by two sensors out of three or from redundancies between sensors, to confirm the presence of the previous vehicle can be done. The data acquired by the sensors C1, C2 and C3 correspond, for example, to the size, the speed and the relative position of the preceding vehicle.

In one embodiment, the preceding vehicle is used as the primary target vehicle for longitudinal guidance purposes and the current vehicle provides a defined inter-vehicular distance by adjusting its speed. In particular, the data transmitted to the vehicle V _k can be:

- acquired by one of the sensors of the preceding vehicle;

 - Acquired by one of the sensors of an element communicating road infrastructure near the current and / or previous vehicle. This is typically a traffic light, a road surface or devices (lines, zebra crossings, etc.) of the road, speed control radars, road signs, etc. ;

 - acquired from connected user devices, such as telephones, tablets, personal assistant or computer. In this situation, the data can be entered by the user or captured by sensors of the user devices (accelerometer, GPS, etc.);

- Informed by an occupant of the previous vehicle and / or current, typically on one of the man-machine interfaces available in these vehicles. More generally, any type of data received via a vehicle communication system G1 or any communication system to any G2 device can be taken into account to determine the reaction of the previous vehicle.

Any type of physical or digital reaction of the previous vehicle can be taken into account. For example, braking, accelerating, changing direction or any change in the trajectory of the previous vehicle, sending an emergency message, changing the driving mode, triggering any alert on the previous vehicle etc. are vehicle reactions.

From this history, a reaction R _n of the preceding vehicle to the object O _n is determined in step S4. If the object O _n is a static object, the determination consists in seeing how the previous vehicle Vu reacted when it was confronted with the object O _n . For example, if the vehicle V _k -ia 4 seconds ahead of V _k , the reaction in the history at t-4 seconds is taken into account.

The object O _n can also be dynamic. In this situation, taking into account the reaction of the preceding vehicle located on an adjacent lane is relevant, especially when the initial detection indicates that the dynamic object is moving towards the current vehicle. Taking into account the sensor data of the previous vehicle is also relevant in this situation.

A lack of reaction of the previous vehicle to an object initially detected or if, conversely, the previous vehicle has reacted without anything has been detected by the current vehicle, a malfunction of the object detection mode is certainly present. In these cases, false-positive or false-negative learning may be carried out, a driver's hand-over instruction may be generated and / or braking may also be implemented for safety. Taking into account the reaction of the vehicle is very relevant in the example mentioned above with reference to step S5 (examples of existence probability intervals). In particular, a particular method taking into account the reaction of the vehicle, a particular embodiment of the present invention, is described below:

A. Method for processing data relating to the assisted driving of a current vehicle (V _k ), for the confirmation of the detection of an object (O _n ) likely to modify the driving of the current vehicle, said current vehicle being preceded by a preceding vehicle (V _k -i), comprising the steps of:

 Acquisition (S1) by at least one sensor (C1; C2; C3) of the current vehicle of data relating to the object;

 In the case where the object is initially detected from said data, execution of the substeps of:

 o determining (S4) a past reaction of the preceding vehicle to the object; confirmation (S5) of the detection of the object according to said determination.

If an object is detected by the sensors of the current vehicle, it is unlikely that the previous vehicle has not responded to this object. Confirming the detection of the object taking into account the reaction of the preceding vehicle substantially improves the reliability of the detection of the object, without the need for additional components (sensor redundancy).

 This is a very good compromise between the reliability of the detection and a complexity / cost control. B. Method according to method A, wherein determining the reaction of the preceding vehicle to the object comprises the sub-steps of: i. acquisition (S3) of a history of reactions of the preceding vehicle; ii. determining (S4), from the history, the past reaction of the preceding vehicle to said object. C. Method according to method B, in which the history is acquired by at least one of the following elements: the sensor of the current vehicle;

at least one sensor of the preceding vehicle; At least one sensor of a road infrastructure element in the vicinity of the current vehicle and / or of said preceding vehicle;

 • data related to at least one of the occupants of the previous vehicle and / or the current vehicle. E. Method according to one of the preceding methods, wherein the reaction of the preceding vehicle is at least one of:

• a braking action;

 • an acceleration action;

 • a modification of the trajectory of the preceding vehicle;

 · The triggering of an object detection alert.

F. method according to one of the preceding methods, wherein the preceding vehicle precedes the current vehicle in one of the following positions: the current vehicle and the preceding vehicle share substantially the same trajectory, the preceding vehicle preceding the current vehicle in the time on said trajectory;

 the preceding vehicle precedes the current vehicle in a queue of vehicles;

 the preceding vehicle precedes the current vehicle on a road lane;

 the preceding vehicle is on a lane adjacent to a lane of a road where current vehicle is located. G. Method according to one of the preceding methods, further comprising the step of:

Generating (S6) a driving assistance instruction according to said detection confirmation.

The instructions are taken in step S6. They can then be directly implemented by a vehicle member (braking action, acceleration, ignition of lights, transmission of an emergency message for example) or notified to another assisted driving module. If no object is finally detected, the instruction does not lead to any particular action. The object O _n having thus been processed, the procedure returns for a new object n + 1 at Step S7.

FIG. 3 represents an exemplary device D of the vehicle V _k . This device D can be used as a centralized device in charge of at least some steps of the method according to the invention, as a remote server or any type of device capable of implementing the steps of the method according to the invention. The data acquisition steps relating to the object are performed by sensors such as sensors C1, C2 and / or C3. The other steps can be performed by the single device D but also be performed for some by other processing devices located at other locations in the vehicle. This device D can take the form of a housing comprising printed circuits, any type of computer or a mobile phone.

The device D comprises a random access memory 1 for storing instructions for the implementation by a processor 2 of the method as described above. The device also comprises a mass memory 3 for storing data intended to be stored after the implementation of the method. The RAM 1 and / or the mass memory 3 store, for example, the reaction history of the preceding vehicle.

The device D may further comprise a digital signal processor (DSP) 4. This DSP 4 receives the data from the sensors to format, demodulate and amplify, in a manner known per se, these data. This treatment can alternatively be performed at the sensors.

The device also comprises an input interface 5 for receiving the data recorded by the sensors and an output interface 6 for transmitting the data of the driving assistance instructions.

The present invention is not limited to the embodiments described above as examples; it extends to other variants. Thus, examples of parameters other than the probability and the number of sensors that can be taken into account for the detection of the object or the determination of the trajectory of the current vehicle are described above. Of course, other parameters such as weather conditions, the detected density of the object or the engine displacement can be taken into account.

Claims

claims 1. Method of processing data relating to the assisted driving of a current vehicle (V _k ), for the detection of an object (O _n ) likely to modify the driving of said vehicle, comprising the steps of:  Acquisition (S1) by at least two sensors (C1; C2; C3) of data relating to said object;  Calculating (S2) a probability of existence of said object, at least from the data of the sensors;  Detection (S5) of said object in function, at least:  o the probability of existence;  o the number of sensors that have acquired, on their own, data indicating the presence of the object;  Generating (S6) a driving assistance instruction according to said detection. 2. Method according to claim 1, comprising, prior to the step of calculating the probability of existence, a step of:  • verification that said object is in the path of said vehicle; in the case where the object is not on the path, generating a non-detection instruction. 3. Method according to claim 2, wherein said object is considered in the path of said vehicle if the object is moving towards the path. 4. Method according to any one of the preceding claims, wherein the detection of the object is also a function of: o the reaction to the object of a vehicle preceding the current vehicle. The method of claim 4, wherein the reaction to said object of the preceding vehicle is determined by implementing substeps of:  i. Acquisition (S3) by the sensors of the current vehicle of a history of reactions of the preceding vehicle;  ii. Determination (S4), from the history, of the reaction of the preceding vehicle to said object. The method of claim 4 or 5, wherein the reaction of the preceding vehicle is at least one of:  · A braking action;  • an acceleration action;  • a change in the trajectory of the current vehicle;  • the triggering of an object detection alert. 7. Method according to one of the preceding claims, further comprising a step of: • determination of a lateral trajectory of the current vehicle; and wherein the object is not detected if said sensors reveal that the object is not in the path of the current vehicle. 8. Computer program comprising instructions for implementing the method according to any one of the preceding claims, when these instructions are executed by a processor. 9. Device (D) for processing data relating to the assisted driving of a current vehicle, for the detection of an object likely to modify the driving of said vehicle, comprising:  At least two sensors (C1, C2, C3) of data relating to said object; A processor (2) arranged to perform the operations of: calculating a probability of existence of said object, at least from the data of the sensors;  o detection of said object in function, at least: ^■ the probability of existence; ^■ the number of sensors that have acquired, on their own, data indicating the presence of the object;  o generating a driving instruction according to said detection. Vehicle (V _R ; V _RI ) having the data processing device according to claim 9.